2012/2/9 Rüdiger Kurz <[email protected]>: > Hi "stanbolers", > > I want to suggest a useful feature for stanbol: > > Coming from the GUI perspective, I want to show related content for the > topic an article deals with. > > Imagine a plain text that will be enhanced by stanbol using dbpedia. The > found entities contain "Steve Jobs" as person and "Apple Inc." as > organization: Now I want to ask stanbol to tell me the topic. One possible > answer would be computer science.
I am currently working on a trainable topic classifier that directly works with the input text (but could be extended in the future to further use more structured input): https://issues.apache.org/jira/browse/STANBOL-197 > In another case the found entities contain > "Steve Jobs" as person and "death" as event: If I ask stanbol again for the > topic a possible answer would be cancer. > > Is that already possible using stanbol, if so: how? Or would that be a > useful feature for stanbol for future. This kind of feature could be implemented with the reasoner components but don't underestimate the effort to write a comprehensive set of rules. -- Olivier http://twitter.com/ogrisel - http://github.com/ogrisel
